Cortexy — AI Safety Platform Redesign

Cortexy — AI Safety Platform Redesign

Role: Designer
Timeline: Jun 2025 – Feb 2026
Scope: Web platform overhaul + mobile app design

———————————————————————————

The Setup

Heavy industries — construction, logistics, mining — have cameras everywhere. But those cameras were just recording. Cortexy was built to change that: AI video analytics that turn existing camera feeds into real-time safety and productivity insights. The tech was powerful. The platform? It was drowning users in data.

The Problem

The legacy platform was built feature-first, design-last. Every new module got bolted on, creating a fragmented experience where critical safety alerts competed with routine data noise. For a Safety Officer managing dozens of live sites, finding a PPE violation buried inside a wall of unfiltered notifications wasn't just frustrating — it was a liability.

Three core problems emerged:

• No design system — each new module broke visual consistency, making scaling impossible

• Cognitive overload — a non-hierarchical stream of alerts meant critical issues got missed

• Data density without clarity — users had to work to extract insight from a platform that should have surfaced it instantly

The Scope

The project covered two parallel tracks:

Web platform — redesigned for three distinct personas: Moderators (oversight), Safety Officers (hazard monitoring), and Coordinators (site logistics). Each persona had different priorities; the design had to serve all three without compromise.

Mobile app — built specifically for field workers who needed quick, at-a-glance safety alerts with zero learning curve. Intentionally limited in scope to prevent information overload on-site.

The Process

A structured Iterative Modular Approach — each module was treated as a standalone design sprint, moving through research → wireframes → stakeholder validation → high-fidelity design → developer handoff → post-dev iteration. We ran 4–5 wireframe rounds per module to validate complex logic (like polygonal zone drawing and directional line-crossing counts) before any pixel-perfect work began.

Module-by-Module Breakdown

Guardian — Safety Monitoring

Pain point: A constant, undifferentiated flood of notifications caused alert blindness. Safety Officers were missing critical violations.
Solution: Rebranded the alerts module as "Guardian" and introduced a tiered severity system — critical violations surface first, minor drifts fall below the fold. 8,000+ alerts now have clear status pathways: Pending → Acknowledged → Compliant/Non-Compliant.

Vehicle Management — Fleet Intelligence

Pain point: High data volumes and missing vehicle identifiers led to misconfigurations and user errors during bulk operations.
Solution: Redesigned the workflow for bulk plate entry and zone assignment with a clear visual hierarchy. Fleet movement, transit trails, and vehicle type breakdowns are now surfaced in a dedicated logistics hub.

Camera Health — CCTV Diagnostics

Pain point: "Dashboard fatigue" — tampering alerts and camera outages were buried alongside routine heartbeat data.
Solution: Progressive disclosure model with colour-coded health markers. A tampered or offline camera is instantly visible without digging through operational data.

SpaceSense — Occupancy Analytics

Pain point: Managing capacity thresholds, dwell times, and live occupancy across multiple cameras simultaneously created a cluttered, unusable interface.
Solution: A unified zone-drawing workflow with a time-slider navigator for scrubbing through historical heatmaps — giving site managers spatial intelligence without the noise.

FlowSense — Traffic Intelligence

Pain point: Configuring virtual line-crossing boundaries on live feeds had no intuitive UI — it required technical workarounds and was prone to error.
Solution: A configuration tool for drawing directional entry/exit lines directly on camera feeds, paired with a global dashboard that converts raw counts into site-level traffic insights.

Image Vault & Insight Vault — Evidence Management

Pain point: Investigators had to manually scroll through massive image databases to find incident evidence — slow, error-prone, and impractical under pressure.
Solution: Advanced date, camera, and alert-type filters, plus a side-by-side image comparison tool to analyze snapshots across different timestamps — turning evidence retrieval from a task into a feature.

Design System

Built on Gluestack UI v2 for scalable core components, with custom-engineered elements for industrial-specific data patterns. Brand colors were refined to meet WCAG accessibility standards throughout.

The Outcome

A unified, modular design system that Cortexy can scale — adding new AI use cases without fragmenting the experience. Complex, multi-layered AI data is now presented through high-contrast dashboards built for instant comprehension, even in high-stakes, time-pressured industrial environments.